Inhalt

[ 993MLPEEAIK19 ] KV (*)Explainable AI

Versionsauswahl
(*) Leider ist diese Information in Deutsch nicht verfügbar.
Workload Ausbildungslevel Studienfachbereich VerantwortlicheR Semesterstunden Anbietende Uni
3 ECTS M2 - Master 2. Jahr Informatik Marc Streit 2 SSt Johannes Kepler Universität Linz
Detailinformationen
Quellcurriculum Masterstudium Artificial Intelligence 2019W
Ziele (*)This course introduces how static and interactive visualization can be facilitated to analyze and better understand AI processes and black-box algorithms during all three phases: model building, model training, and model usage.
Lehrinhalte (*)
  • Visualization Techniques and Tools for AI
  • Visualization Support in Deep Learning
  • Supporting Interpretability & Explainability through Visualization
  • Debugging & Improving Models Using Visualization
  • Comparing & Selecting Models Using Visualization
  • Visualizing Network Architectures, Learned Model Parameters (Edge Weights, Convolutional Filters), Computational Units (Activations, Gradients for Error Measurement), Neurons, Aggregated Information
  • Case Studies and Selected Research
Beurteilungskriterien (*)Written exam (oral exam in exceptional cases) combined with practical exercises.
Lehrmethoden (*)Slide presentation with case studies, tutorials, in-class exercises, and practical project activities.
Abhaltungssprache Englisch
Lehrinhalte wechselnd? Nein
Präsenzlehrveranstaltung
Teilungsziffer -
Zuteilungsverfahren Direktzuteilung